Internet Explorer seems to have trouble with pages served using
the gzip HTTP transfer-encoding. Earlier versions (5.0) would
sometimes crash, while IE6 doesn't crash but does go into some
kind of weird state. Here are the IE6 symptoms:
- Page does not load
- The browser does not even bother to draw in the web page area
of the window (meaning that if you put another window on top of
it, and then bring IE back to the foreground, it will still show
the old window or whatever other junk was on the screen). It
does however draw its menues and toolbars.
- Hitting reload usually doesn't help, although occasionally it
will and the page will load correctly.
- Navigating to another site does work, and seems to bring the
browser "back to life".
- The IE icon will sometimes spin as though it is loading a page,
however a 'netstat' on the server reveals that there aren't
any open connections.
- After IE goes into the weird state, the most reliable way
to make it work again is to restart IE. Waiting a long time
also seems to help though.
My suspicion is that there is a bug in IE's networking code that
causes the connection to get into some invalid state.
This problem seems to appear in conjunction with several aspects
of the page being loaded (but I'm not certain of any of these):
- Multiple simultaneous loads (as with frames)
- iframes
- javascript
- utf8
- chunking
- https (seems to make the problem occur more often)
- short http keep-alive timeout (long keep-alive makes the problem
mostly go away, while short or no keep-alive makes it more likely)
I want a better understanding of the problem and the ability to
reliably avoid it while still using gzip and the same page structure.
What I need from you is:
- A simple demonstration of the problem. This can just be a little
python server, or whatever. It should run in linux. The demo
_must_ demonstrate the problem on IE6, though it would be nice
if it also applied to IE 5.0 and 5.5.
- A fix for the problem that involves changing something on the
server. The simple demo should also demo this fix. Switching off
gzip, using a super-long keepalive, making a major change to the
page content, or switching to a different browser is not an
acceptable fix.
The web application experiencing the problem is internal and
therefore I can't provide it as a demonstration. I've experienced
the problem with several unrelated applications though, so it's
not completely peculiar to my current app. |
Request for Question Clarification by
joseleon-ga
on
30 Sep 2003 02:08 PDT
Hello, beavis:
Which webserver are you using? Are you using Apache+mod_gzip? Which version?
Regards.
|
Request for Question Clarification by
joseleon-ga
on
08 Oct 2003 01:52 PDT
Hello, beavis:
Now the question has been reopened I will continue researching and
be sure I won't answer until you get what you want. Is there any way I
can test the problems you say? That is, is that server live on the
internet so I can access to it?
Regards.
|
Clarification of Question by
beavis-ga
on
08 Oct 2003 16:54 PDT
Unfortunately the server cannot be made publicly available. It uses
zlib to perform the compression though, so hopefully it's not too
unusual. I should reemphasize that this problem does not occur on all
gzipped pages. It only sometimes occurs, and I suspect that it may be
related to the http chunk boundaries. The specific page layout is a
frameset with a several html pages, one of which includes an external
css page. From what I can tell, IE is getting hung up on the css.
|
Request for Question Clarification by
joseleon-ga
on
09 Oct 2003 00:16 PDT
Hello, beavis, a few more questions:
Is there any possibility you post here a copy of the data is sent to
the browser, that is, served by the webserver? The data must be still
compressed and I will try to perform some tests with that data. Bear
in mind that random problems like you describe are very hard to debug.
The server are you using implements HTTP transfer-enconding, but
what RFC have you followed to implement it? That is, how you could be
100% sure that there is not a bug on the server?
Regards.
|
Request for Question Clarification by
joseleon-ga
on
15 Oct 2003 23:41 PDT
Hello, beavis:
Are you still interested to find a solution to this problem? Please, let me know.
Regards.
|
Request for Question Clarification by
joseleon-ga
on
25 Oct 2003 01:22 PDT
Hello, beavis:
Any updates on this matter? Could I help you?
Regards.
|
